|Publication number||US7643658 B2|
|Application number||US 11/040,587|
|Publication date||Jan 5, 2010|
|Filing date||Jan 21, 2005|
|Priority date||Jan 23, 2004|
|Also published as||EP1566788A2, US20050197923|
|Publication number||040587, 11040587, US 7643658 B2, US 7643658B2, US-B2-7643658, US7643658 B2, US7643658B2|
|Inventors||Andrew Roger Kilner, Victoria Sophia Jennings, Sebastian Aleksander Paszkowicz, Robert Mark Stefan Porter, Ratna Rambaruth, Clive Henry Gillard|
|Original Assignee||Sony United Kingdom Limited|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (29), Non-Patent Citations (1), Referenced by (65), Classifications (26), Legal Events (3)|
|External Links: USPTO, USPTO Assignment, Espacenet|
The present application is based on UK Application No. 0412037.4, filed on May 28, 2004 and UK Application No. 0401492.4, filed on Jan. 23, 2004, which are each incorporated herein by reference in their entirety.
1. Field of the Invention
This invention relates to displays.
2. Description of the Prior Art
Electronic displays are commonly used in shops and other public areas to carry advertising material and the like. Such an arrangement may be referred to as “digital signage”.
A common arrangement is for a television monitor and video player (e.g. a VHS cassette tape player) to play out material encouraging shoppers to buy goods laid out near to the television monitor. A more advanced arrangement is provided by, for example, the SonyŽ NSP100™ digital playout system. This device is a networked playout device which receives video (and, optionally, audio) material and a playout schedule over a digital network, and then plays out the video and audio material on a display device at times indicated by the schedule.
While these arrangements provide a degree of user control over the material played out—for example, the user can select a tape cassette or can load desired material onto the NSP100 playout system—it is relatively difficult to tailor the displayed material to best suit a current audience.
WO 01/45004 and U.S. Pat. No. 5,966,696 disclose digital signage arrangements which use a detection of the audience to affect the material played out. In U.S. Pat. No. 5,966,696, proximity detectors are used to detect the presence of a viewer at a display screen, which causes the display screen to be switched from an idle to an active mode. WO 01/45004 purports to use a detection of certain characteristics of a viewer in front of a display screen to control the selection of material to be played out at that display screen.
This invention provides a display arrangement comprising:
an image display device having a first camera directed towards positions adopted by users viewing the display;
one or more further cameras directed towards other respective locations;
a face detector for detecting human faces in images captured by the cameras; and
means for detecting those faces which appear in images captured by both the first camera and at least one of the further cameras.
The invention provides an improvement in, for example, the field of digital signage, by correlating a face detection of a viewer observing a display screen—which might, for example, be displaying advertising material—with a detection of that same viewer by a camera at another location. The other location might be, for example, a shop checkout, or a place in a shop where certain goods (e.g. the goods relating to the displayed advertising material) are arranged for user purchase. This arrangement can thus give improved feedback to the originators of the advertising material as to the success of that material in persuading consumers to buy a product.
Various other respective aspects and features of the invention are defined in the appended claims.
The above and other objects, features and advantages of the invention will be apparent from the following detailed description of illustrative embodiments which is to be read in connection with the accompanying drawings, in which:
The camera 110 is integral to the mounting of the plasma display screen 130 and is used to capture images of human faces as they view the display screen. The captured video images are analysed to detect human faces and to perform facial recognition analyses. The camera 110 is discreetly mounted, perhaps in a frame of the display, so as not to distract the viewer's attention from the promotional display presentation comprising advertising images for products or services available and so as not to inhibit a viewer's normal reaction to the displayed material.
The plasma screen 130 is driven by the graphics card 120 and presents display information from a selection of possible promotional display sequences stored on the data storage device 190. The processor 194 is (or is based on) a Sony™ NSP100™ processor and is operable to control download of display sequences via the network 196. The basic NSP100 processor generally operates in a “push” mode, which is to say that it receives content pushed to it by a server, and displays that content. Here, the processor 194 can also operate in a “pull” mode by which it requests or downloads appropriate content for display, for example in dependence on the nature of the viewing audience.
The control computer 600 can configure the local display arrangement, the particular configuration determining the selectable range of display sequences available at the particular plasma display screen 130.
The graphics card 120 comprises an MPEG (or a proprietary format used in the NSP100) decoding unit that facilitates display of stored MPEG format display material on the plasma screen.
The audio generator 140 co-operates with the graphics card to produce sound effects for the promotional display sequence. The sound effects may be reproduced either with an MPEG video sequence or independently, as required for a particular application.
The face detection/recognition module 150 comprises face detection software to identify the presence of human faces that appear within a predetermined spatial range of the camera (and associated display screen) and face recognition software. The face recognition software is operable to recognise individual faces previously detected at the same camera, at other camera locations or stored within a database of detected faces. The face detection algorithm works by analysing the field of view for shapes that could be human faces and searching for matches of features of a detected object with template facial features stored in memory. Facial movements such as eye-blinks and mouth movements can also be used to determine whether face-like objects indeed correspond to human faces.
A set of characteristic facial parameters is established by analysis of the colour darkness levels of image pixels. Facial features such as eyes, eyebrows, cheekbones and noses can be identified by searching for abrupt changes in darkness levels across adjacent image regions. Colour histograms of image pixels can be used to detect skin tone and hair colour. A plurality of anchor points on a detected face can be defined to provide a representation of facial bone structure. The anchor points are determined from changes in darkness levels. The anchor points typically include the corners of the mouth, points on the brow line, cheek bones and points corresponding to the bridge and the tip of the nose. The anchor points are joined to form a mesh of triangles and the angles of each triangle are used as a characteristic set of facial parameters that can be used to match similar detected faces and thus to perform facial recognition analysis.
In an alternative arrangement, a so-called eigenface method of face recognition is used. According to the eigenface method the face as a whole is examined, rather than local facial features. A database of detected faces is first analysed to define a composite face and each newly detected face is compared as a whole to the composite face. A numerical value specifying differences between a detected face and the composite face is used as the characteristic facial parameter data.
A specific example of face detection, tracking and similarity detection will be described below with reference to
The skilled man will appreciate that face detection techniques and the like do not tend to give an absolute answer, but rather a probability that an image area represents a face, or that a set of images from two cameras contain the same face, and so on. The probabilities can be compared with a threshold to give a result which appears to be a clear “yes” or “no”, but its derivation will still be via a statistical process.
The statistical analysis module 160 maintains cumulative statistical data on the human faces detected locally by the camera 110 including a headcount of distinct faces, gender data, age category data and time spent by each detected face within a predetermined spatial range of the plasma display screen 130 (i.e. the in-shot time). The face detection/recognition module 150 is operable to detect both the in-shot time of each human face and the dwell time, which is the time for which the detected human face directly views (i.e. is oriented directly towards) the display screen 130. A cumulative dwell time may also be recorded. The cumulative dwell time represents the total time spent either by all detected faces within a given time period or by a predetermined category of detected faces (e.g. male faces or children's faces) directed towards the display screen associated with the particular camera). Alternatively, a cumulative dwell time can be calculated for a given face ID or for a given category of faces such that it represents the total dwell time at a plurality of different display screens/cameras at different locations in the network. The local memory 192 is available for local data processing and for storage of a limited quantity of local statistical data prior to the downloading of that data for a global statistical analysis by the control computer 600.
It will be appreciated that display technologies other than plasma screens can be used. It will also be appreciated that within the context of a store-wide network, the skilled person can select which processing operations should take place at a display terminal and which should take place at one or more central servers. Indeed, some operations may take place at both, depending on the nature of the operation. For example, some statistical operations used to determine which content should be displayed may be best carried out quickly and locally at the terminal. A more detailed analysis of the same information, for example to harvest data for marketing advertising space on the displays, could be carried out at a server.
The images at 212 were captured two minutes ago; the images at 214 were captured four minutes ago and the images at 216 were captured 6 minutes ago as indicated by the grid column headings. Each row of the grid represents images presented to a given camera 110. A GUI panel 220 is used to display a number of most recently logged faces. The number of faces displayed in this GUI panel is selectable by a user highlighting the appropriate button 222, 224, or 226 on the display screen corresponding respectively to one, two and four images. The GUI 200 also comprises a face statistics panel 230 associated with a user-selected one of the representative keystamp images displayed in the recently logged faces GUI panel 220. A representative keystamp image 238 with which the statistics are associated is also displayed within the face statistics panel 230. In this example arrangement the statistics displayed are a face ID 232, a cumulative dwell time 234 of detected faces at the given camera, which in this example is seven seconds and the in-shot time 236, which is three seconds. The cumulative dwell time in this case represents the cumulative time spent by all detected faces (of any category) oriented directly towards the plasma display screen.
At a stage 310 a captured image sequence is analysed locally by the face detection/recognition module 150. The analysis may be at least partially performed before image capture associated with the detected face has terminated.
At a stage 320 a local face ID is assigned to the captured image sequence (or to one face in the sequence) in dependence upon predetermined characteristics of the detected human face.
Next, at a stage 330, the statistical analysis module 160 determines the gender category (i.e. male/female) of the detected face based on information derived from the face detection/recognition module 150 and at a stage 340 the detected face is placed in one of a number of predetermined or variable age categories: such as child; adolescent; working-age adult; retired adult. In general, the age categories may well be decided by the marketing requirements of the products on sale. The age category may be determined by the face detection/recognition module 150 from an analysis of the skin tone/texture, hair colour and the relative proportions of the facial features.
At a stage 350, the gender category and age category associated with the local face ID and additional information including the current day and time, the in-shot time and dwell time are assimilated as a customer profile and this information is stored to local memory 192 (and optionally to remote memory) at a stage 360.
In alternative arrangements additional information such as ethnic origin, specific physical characteristics and categories of emotional expressions can be used to establish or augment the customer profile.
Next, at a stage 370, one of a plurality of locally stored promotional sequences is selected for display on the plasma screen 130 in dependence upon the customer profile that was determined at the stage 350. The promotional sequence is selected such that it is specifically targeted at the particular customer profile.
At a stage 380 the selected promotional display sequence and any accompanying audio data are retrieved from the local data storage device 190, decoded by the graphics card 120 and the audio generator and presented to the customer using the plasma display screen 130.
At a stage 390, when the selected advertising sequence has been displayed once from beginning to end, it is detected via the camera 110 and face detection/recognition module 150 whether the same human face is still in shot. If not, and no other face is detected in shot then an idle timer is triggered at a stage 392 to monitor the time elapsed between detection of individual human faces. However, if the human face having the same face ID allocated at the stage 320 is still in shot at the stage 390 then the process proceeds to a stage 394 whereupon either an extended advertising sequence directed to the same product as the initial advertising sequence is displayed, or a newly selected promotional sequence that is also appropriate to the customer profile is displayed on the plasma screen 130.
At a stage 410 a face is detected by a camera, which is associated with a plasma display screen 110 located in the audio/video sales section of the store.
At a stage 420 the face detection/recognition module determines if the detected face is the face of a child and, if so, displays a promotion for a cartoon digital video disc (DVD) offer. However, if at the stage 420 it is determined that the detected face does not belong to a child the process proceeds to a stage 430, where the face detection/recognition module determines whether the detected face is likely to belong to the over sixty-five age group. If the face is in fact categorised as over sixty-five then the audio generator 140 is controlled to set a louder than normal audio volume at a stage 432 and subsequently at a stage 434 a classic film video promotional sequence is displayed on the plasma screen 110.
If at the stage 430 it is determined that the detected face does not belong to the over sixty-five age group then it is determined at a stage 440 whether or not the detected face is male. If the face is in fact male the process proceeds to a stage 450 whereupon a science-fiction DVD offer is displayed on the plasma screen 130. During display of the science-fiction promotional sequence, the camera 110 continues to monitor (at a stage 452 ) the male face and determines whether the facial expression can be categorised as one of boredom at any point during the display of the promotional material or whether the detected face begins changes orientation such that it is no longer directly viewing the display screen. If no such detection is made then the sci-fi DVD promotion continues to be displayed to the customer. However, if a bored face is in fact detected or if the face turns away from the display by more than a predetermined angle then the process proceeds to a stage 454 and a different promotional sequence is displayed on the display screen, in this case a promotion for an action/adventure DVD offer, in an attempt to regain the attention of the viewer and increase the cumulative dwell time.
If at a stage 440, the detected face was determined to be female then at a stage 460 a period drama DVD promotional sequence is launched on the plasma display screen 130. During display of the period drama promotional sequence, the camera 110 continues to monitor (at a stage 462) the female face and determines whether the facial expression can be categorised as one of boredom at any point or whether the detected face begins to change orientation such that it is no longer directly viewing the display screen. If no such detection is made then the period drama promotion continues to be displayed to the female customer. However, if a bored face is in fact detected or if the face turns away from the display screen then the process proceeds to stage 464 and a promotional sequence for a romantic comedy DVD is instead presented on the plasma display screen 130.
The flow chart of
The display screen 130 is initially in an inactive mode, which is a power-saving state but the camera 110 and face detection/recognition module 150 are still powered-up and fully operational. At stage a 510 the face detection/recognition module detects the presence of a human face at camera 1 and at a stage 520 the statistical analysis module 160 logs the time of first detection of the human face.
At a stage 530 the processor 194 sends a control signal to the plasma display screen 130 to initiate an active mode in which full power is supplied to the screen and display of a promotional image/audio sequence is initiated.
At a stage 540 the statistics analysis module 160 increments the local head count for camera 1. Prior to incrementing the local head count a cross-check is performed, by the face detection/recognition module 150, of recently detected faces to determine if the currently detected face has been recently logged at the same camera and, if this is the case, then the local head count is not incremented. The time window over which this cross-check is performed in this particular arrangement is a 2 hour period, but the period is configurable by an operator via the control computer 600.
At a stage 550, a subset of images of the human face captured by the camera is stored either in the memory 192 or on the data storage device 190 for subsequent viewing and/or analysis. The images are captured at a predetermined frequency and quality that is set in dependence upon the available storage capacity.
At a stage 560 the face detection/recognition module 150 determines whether or not the human face is still in shot. If the face is determined to be still in shot then the image capture continues at a stage 562. However, if it is determined that the human face is not in shot then an in-shot timer is stopped and at a stage 570 the captured image sequence is analysed by the face detection/recognition module 570 to detect the total dwell time of the identified human face at camera 1. The total dwell time is the time that the face was oriented directly towards the camera (within a given tolerance margin) and hence the display screen.
At a stage 570 the cumulative dwell time (for customers of all face categories) is also incremented to include the dwell time of the currently detected face. At a stage 580, when it has been determined that the human face is not longer in shot of camera one and that no other human face is in shot, an idle time counter is started. While the idle time counter is running it is determined whether the maximum idle time, which in this case is set to be five minutes, has elapsed. During this time the display screen remains in active mode, but displays a static image rather than an animated sequence. When the idle time exceeds the maximum idle time without a further human face having been detected by the face detection/recognition module 150, the processor 194 powers down the plasma display screen into an inactive mode at a stage 592 to reduce power consumption.
In this arrangement the idle time counter begins to count as soon as it is determined that there is no human face in shot, but in alternative arrangements (such as that described above with reference to
In this particular arrangement a control computer 600 is provided to collate statistics from individual cameras, to store a historical database of detected faces and to perform an overall statistical analysis of detected human faces at the plurality of locations. Alternative arrangements perform the overall analysis in a distributed manner without the need for a control computer 600, for example using a “grid” type computer system.
The control computer comprises: a statistical analysis unit 610 having a global headcount and demographics analyser 612, a face correlation module 614, a customer path-tracking module and a purchase/advert correlator; a statistics database 620; a display sequence repository 630; a local display configuration module 640; an input device 650 and a network interface 660 which, amongst other functions, can provide access to a remote storage facility.
The purchase/advert correlator may be linked to a card reader (not shown) which can read customer loyalty cards and detect data stored on or by them. The card reader may be disposed at the store's checkout.
The control computer 600 is operable to download stored statistics associated with local display screens and associated cameras across the network 196 and to collate and analyse those statistics before storing them in the statistics database 620. The statistics database 620 stores data covering a longer timespan (in this example statistics for the full year) than the statistics stored locally in memory 192 at the display screens. Representative keystamp images and a characteristic set of image parameters for each detected face are also stored in memory at the control computer. The display sequence repository 630 stores a plurality of different promotional display sequences for selective downloading to individual plasma display screens. The local display configuration module 640 enables configuration of each local display, either automatically under the control of a software program or manually by an operator of the control computer using the input device 650.
In the statistical analysis unit 610, the global headcount and demographics analyser 612 collates headcount statistics from each of the local display arrangements thus provides a comparison of the number of viewers attracted to each display (see the description of
The face correlation module 616 collates data on the local face IDs from individual local display arrangements and performs comparisons of characteristic face parameters determined by the face detection/recognition module 150 to determine matches for a given detected face between different cameras. Accordingly, local face IDs are correlated and assigned a global face ID where a match is found for a given face at two or more cameras. For a given global face ID, detection times and dwell times at each different camera are collated to track the progress of a given customer through the store and to establish the time spent viewing each display screen. This information is used by the customer path-tracking module 616 to determine the most likely paths through the store taken by a given subset of consumers on a given day of the week and at a given time of day, or at a given season, or in a particular purchasing “environment”—such as during a major sporting event, on a rainy afternoon, on a summers day etc (see description of
The purchase/advert correlator 618 links purchases of products promoted on the display screens to viewing events at the plasma display screens 130. The purchase information is derived from the product barcode that is scanned at the checkout. A camera located at the checkout detects the purchasing customer's face and attempts to perform a match with a face recently detected at the relevant plasma display screen. If the purchasing customer's face matches that of one of the faces detected on a plasma display screen on which a relevant promotional sequence was actually viewed then a correlation between the advertisement and the purchase is logged along with the dwell time. Information from cameras that are directed towards the product shelves is also used to determine whether or not a promoted product was actually purchased by the customer associated with a face ID detected viewing the relevant plasma display screen.
The purchase/advert correlator 618 is also operable to analyse information on customer loyalty cards. A customer loyalty card is used to store information with regard to the purchasing profile and/or demographic grouping of the particular individual. Customer details stored on the loyalty card can thus be correlated with face detection/recognition information from cameras throughout the store and with information on which advertised products the customer has actually purchased. This provides an effective way of assessing the impact of advertising on an individual customer's purchasing behaviour. The correlation of the information from the checkout (including for example one or more of the purchased product barcode, customer facial characteristics and customer loyalty card data) with faces detected at the local display screens in this way facilitates an objective assessment of the effectiveness of the promotional displays.
Apart from the cameras that are integral to the plasma display screens, other cameras (not shown) are also provided at the store checkouts (or indeed at other locations within the store). These checkout cameras are directed towards the purchasing customer for the purpose of detecting the customers' faces to determine whether a customer was previously detected viewing a promotional sequence on one of the plasma display screens. Furthermore, some cameras may be directed towards the goods for sale rather than towards the customer, which allows customer behaviour in terms of picking up goods from the shelf to be monitored and correlated with the face detection/recognition data from other cameras.
The user can access local statistics associated with each camera displayed on the GUI by selecting that camera using an input device such as a computer mouse. Each camera has a pull-down menu of associated local statistics including local headcount and dwell time (cumulative, by customer or otherwise). For a particular display area the headcount and dwell time will be functions of both the product on sale at the corresponding location and the promotional media being displayed at that display device. In the GUI the headcounts may be represented by colour coding e.g. by using red to identify headcount “hotspots” and blue to represent headcount “coldspots” (places with relatively low headcount) on the location map.
At a stage 1100, the control computer downloads, from the network, locally stored statistics for face detections at each of the plurality, N, of cameras situated at different locations throughout the store. The downloaded information comprises: local face IDs; dates and times of first detection for each local face ID; dwell time for each face ID at the particular camera; a set of face matching parameters used for cross-correlation of faces detected at different cameras at different times; and product IDs of adverts displayed to a customer having a given local face ID, which are used to correlate product sales with promotional sequences viewed on the plasma display screens 130.
At a stage 1200, the local headcount statistics of individual cameras are used to update the global headcount statistics (as illustrated in
The process then proceeds to a stage 1420 where the path of the customer having a given global face ID is tracked between cameras at different locations in the store using the times at which the face was first detected at each camera and dwell times at individual display screens 130. At a stage 1430 a cumulative dwell time for this particular customer that includes the dwell time at every camera location at which their face was detected is calculated.
A further sequence of processing operations is performed if it is determined at a stage 1400 that a local face ID corresponds to a face detected at two or more cameras. In particular, at a stage 1450 it is determined whether one of the cameras at which the face detection was made was a camera directed to a store checkout and if so, the global face ID is associated at a stage 1460 with product advert IDs displayed at the display screen(s) at which the face corresponding to that global face ID was detected. At a stage 1470 the goods purchased by the customer at the checkout, as identified by product barcodes, are correlated with the product advert IDs and finally at a stage 1480 the correlation between purchased products and viewed adverts is logged in the statistics database 620 along with the dwell time at the display screen on which the relevant promotional advert was viewed.
The arrangements described here may enable the effectiveness of advertisements to be determined. The measure of effectiveness could be used to calculate an appropriate fee for the placement of advertisements on the display devices. Such digital signage solutions may be suitable not only for store (shop) environments but for any suitable advertising locations such as (for example) entertainment venues or railway/bus stations. Further, the techniques are not limited to advertising but may also be applied to monitoring the effectiveness of distribution and tailoring of information displayed in general. The techniques could find application in museums, schools, waiting areas etc.
In this example the upper left-hand box in the grid 2100 has been selected so statistics for January 25th from 9:00 am until 10:00 am are presented. The statistics displayed comprise the total number of faces detected on all cameras in that time period, the breakdown of male and female faces, the average dwell time and the average in shot time.
The description of
First, a face detection technique will be described.
Then, a face “tracking” technique, operable to associate detected faces or face positions in a group of successive images, will be described.
Finally, a face “similarity” technique will be described. This technique is operable to associate together two or more distinct face “tracks”, either from the same camera or from different respective cameras.
The example face detection technique is arranged as two phases.
Unlike some previously proposed face detection methods, the present method is based on modelling the face in parts instead of as a whole. The parts can either be blocks centred over the assumed positions of the facial features (so-called “selective sampling”) or blocks sampled at regular intervals over the face (so-called “regular sampling”). The present description will cover primarily regular sampling, as this was found in empirical tests to give the better results.
In the training phase, an analysis process is applied to a set of images known to contain faces, and (optionally) another set of images (“nonface images”) known not to contain faces. The process can be repeated for multiple training sets of face data, representing different views (e.g. frontal, left side, right side) of faces. The analysis process builds a mathematical model of facial and nonfacial features, against which a test image can later be compared (in the detection phase).
So, to build the mathematical model (the training process 10310 of
One such histogram is prepared for each possible block position, by repeating the above steps in respect of a large number of test face images. So, in a system which uses an array of 8×8 blocks, 64 histograms are prepared. In a later part of the processing, a test quantised attribute is compared with the histogram data; the fact that a whole histogram is used to model the data means that no assumptions have to be made about whether it follows a parameterised distribution, e.g. Gaussian or otherwise. To save data storage space (if needed), histograms which are similar can be merged so that the same histogram can be reused for different block positions.
In the detection phase, to apply the face detector to a test image 10350, successive windows in the test image are processed 10340 as follows:
As mentioned above, in the training phase, a set of “nonface” images can be used to generate a corresponding set of “nonface” histograms. Then, to achieve detection of a face, the “probability” produced from the nonface histograms may be compared with a separate threshold, so that the probability has to be under the threshold for the test window to contain a face. Alternatively, the ratio of the face probability to the nonface probability could be compared with a threshold.
Extra training data may be generated by applying “synthetic variations” 10330 to the original training set, such as variations in position, orientation, size, aspect ratio, background scenery, lighting intensity and frequency content.
A face tracking algorithm will now be described. The tracking algorithm aims to improve face detection performance in image sequences.
The initial aim of the tracking algorithm is to detect every face in every frame of an image sequence. However, it is recognised that sometimes a face in the sequence may not be detected. In these circumstances, the tracking algorithm may assist in interpolating across the missing face detections.
Ultimately, the goal of face tracking is to be able to output some useful metadata from each set of frames belonging to the same scene in an image sequence. This might include:
The tracking algorithm uses the results of the face detection algorithm, run independently on each frame of the image sequence, as its starting point. Because the face detection algorithm may sometimes miss (not detect) faces, some method of interpolating the missing faces is useful. To this end, a Kalman filter is used to predict the next position of the face and a skin colour matching algorithm was used to aid tracking of faces. In addition, because the face detection algorithm often gives rise to false acceptances, some method of rejecting these is also useful.
The algorithm is shown schematically in
In summary, input video data 10545 (representing the image sequence) is supplied to a face detector of the type described in this application, and a skin colour matching detector 10550. The face detector attempts to detect one or more faces in each image. When a face is detected, a Kalman filter 10560 is established to track the position of that face. The Kalman filter generates a predicted position for the same face in the next image in the sequence. An eye position comparator 10570, 10580 detects whether the face detector 10540 detects a face at that position (or within a certain threshold distance of that position) in the next image. If this is found to be the case, then that detected face position is used to update the Kalman filter and the process continues.
If a face is not detected at or near the predicted position, then a skin colour matching method 10550 is used. This is a less precise face detection technique which is set up to have a lower threshold of acceptance than the face detector 10540, so that it is possible for the skin colour matching technique to detect (what it considers to be) a face even when the face detector cannot make a positive detection at that position. If a “face” is detected by skin colour matching, its position is passed to the Kalman filter as an updated position and the process continues.
If no match is found by either the face detector 10450 or the skin colour detector 10550, then the predicted position is used to update the Kalman filter.
All of these results are subject to acceptance criteria (see below). So, for example, a face that is tracked throughout a sequence on the basis of one positive detection and the remainder as predictions, or the remainder as skin colour detections, will be rejected.
A separate Kalman filter is used to track each face in the tracking algorithm.
It is noted that the tracking process is not limited to tracking through a video sequence in a forward temporal direction. Assuming that the image data remain accessible (i.e. the process is not real-time, or the image data are buffered for temporary continued use), the entire tracking process could be carried out in a reverse temporal direction. Or, when a first face detection is made (often part-way through a video sequence) the tracking process could be initiated in both temporal directions. As a further option, the tracking process could be run in both temporal directions through a video sequence, with the results being combined so that (for example) a tracked face meeting the acceptance criteria is included as a valid result whichever direction the tracking took place.
Advantages of the Tracking Algorithm
The face tracking technique has three main benefits:
In this example, the images 10810 contain one or more faces. In particular all of the images 10810 in the scene include a face A, shown at an upper left-hand position within the schematic representation of the image 10810. Also, some of the images include a face B shown schematically at a lower right hand position within the schematic representations of the images 10810.
A face tracking process is applied to the scene of
In the system described above and in PCT/GB2003/005186, during face detection and tracking, a person's track is terminated if the face turns away from the camera for a prolonged period of time or disappears from the scene briefly. On returning to the scene, the face is detected again but a different track is started, and the new track is given a different identification (ID) number.
So-called “face similarity” or “face matching” techniques will now be described.
The aim of face similarity is to recover the identity of the person in these situations, so that an earlier face track and a later face track, or respective tracks from different cameras (relating to the same person), may be linked together. In this arrangement, at least in principle, each person is assigned a unique ID number. When the person returns to the scene, the algorithm attempts to reassign the same ID number by using face matching techniques.
The face similarity method is based on comparing several face “stamps” (images selected to be representative of that tracked face) of a newly encountered individual to several face stamps of previously encountered individuals or individuals encountered elsewhere. Note that face stamps need not be square. Several face stamps belonging to one individual are obtained from the face detection and tracking component of the system. As described above, the face tracking process temporally links detected faces, such that their identity is maintained throughout the sequence of video frames as long as the person does not disappear from the scene or turn away from the camera for too long. Thus the face detections within such a track are assumed to belong to the same person and face stamps within that track can be used as a face stamp “set” for one particular individual.
A fixed number of face stamps is kept in each face stamp set. The way in which face stamps are selected from a track is described below. Then, a “similarity measure” of two face stamp sets will be described. There will then follow a description of how the similarity method is used within the face detection and tracking system. But first,
At a first stage 12300, so-called “area of interest” logic derives those areas within an image at which face detection is to take place. At those areas of interest, face detection 12310 is carried out to generate detected face positions. Then face tracking 12320 is carried out to generate tracked face positions and IDs. Finally, the face similarity function 12330 is used to match face stamp sets.
Selection of Stamps for the Face Stamp Set
In order to create and maintain a face stamp set, a fixed number (n) of stamps is selected from a temporally linked track of face stamps. The criteria for selection are as follows:
If fewer than n stamps are gathered for one face stamp set, this face stamp set is not used for similarity assessment as it probably does not contain much variation and is therefore not likely to be a good representation of the individual.
This technique has applications not only in the face similarity algorithm, but also in selecting a set of representative pictures stamps of any object for any application.
A good example is in so-called face logging. There may be a requirement to represent a person who has been detected and logged walking past a camera. A good way to do this is to use several pictures stamps. Ideally, these pictures stamps should be as different from each other as possible, such that as much variation as possible is captured. This would give a human user or automatic face recognition algorithm as much chance as possible of recognising the person.
In comparing two face tracks, to detect whether they represent the same individual, a measure of similarity between the face stamp set of a newly encountered individual (setB) and that of a previously encountered individual (setA) is based on how well the stamps in face stamp setB can be reconstructed from face stamp setA. If the face stamps in setB can be reconstructed well from face stamps in setA, then it is considered highly likely that the face stamps from both setA and setB belong to the same individual and thus it can be said that the newly encountered person has been detected before.
The same technique is applied to the arrangement described above, namely the selection of face images for use as a face stamp set representing a particular face track. In that case, the similarity between each newly encountered candidate face stamp and the existing stamps in the set, and the mutual similarity between stamps within the existing set, would be considered in the same way as a the similarity between a stamp from setB and stamps from setA in the description that follows.
A stamp in face stamp setB is reconstructed from stamps in setA in a block-based fashion. This process is illustrated schematically in
Each non-overlapping block 2050 in the face stamp 12040 is replaced with a block chosen from a stamp in face stamp setA. The block can be chosen from any stamp in setA and from any position in the stamp within a neighbourhood or search window 12100 of the original block position. The block within these positions which gives the smallest mean squared error (MSE) is chosen to replace the block being reconstructed by using a motion estimation method. (A good motion estimation technique to use is one which gives the lowest mean squared error in the presence of lighting variations while using a small amount of processing power). Note that the blocks need not be square. In the example shown, a block 12060 is replaced by a nearby block from the stamp 12000; a block 12070 by a block from the face stamp 12010; and a block 12080 by a block from the face stamp 12020, and so on.
When a face stamp is being reconstructed, each block can be replaced by a block from a corresponding neighbourhood in the reference face stamp. But optionally, in addition to this neighbourhood, the best block can also be chosen from a corresponding neighbourhood in the reflected reference face stamp. This can be done because faces are roughly symmetrical. In this way, more variation present in the face stamp set can be utilised.
Each face stamp used is of size 64×64 and is divided into blocks of size 8×8. The face stamps used for the similarity measurement are more tightly cropped than the ones output by the face detection component of the system. This is in order to exclude as much of the background as possible from the similarity measurement.
To crop the image, a reduced size is selected (or predetermined)—for example 50 pixels high by 45 pixels wide (allowing for the fact that most faces are not square). The group of pixels corresponding to a central area of this size are then resized so that the selected area fills the 64×64 block once again. This involves some straightforward interpolation. The resizing of a central non-square area to fill a square block means that the resized face can look a little stretched.
The choice of a cropping area (e.g. a 50×45 pixel area) can be predetermined or can be selected in response to attributes of the detected face in each instance. Resizing in each case to the 64×64 block means that comparisons of face stamps—whether cropped or not—take place at the same 64×64 size.
Once the whole stamp is reconstructed in this way, the mean squared error between the reconstructed stamp and the stamp from setB is calculated. The lower the mean squared error, the higher the amount of similarity between the face stamp and face stamp setA.
In the case of a comparison between two face stamp set, each stamp in face stamp setB is reconstructed in the same way and the combined mean squared error is used as the similarity measure between the two face stamp sets.
Thus the algorithm makes full use of the fact that several face stamps are available for each person to be matched. Furthermore the algorithm is robust to imprecise registration of faces to be matched.
In the system described above, newly gathered face stamp sets are reconstructed from existing face stamp sets in order to generate a similarity measure. The similarity measure obtained by reconstructing a face stamp set from another face stamp set (A from B) is usually different from when the latter face stamp set is reconstructed from the former one (B from A). Thus, in some cases, an existing face stamp set could give a better similarity measure when reconstructed from a new face stamp set than vice versa, for example if the existing face stamp set were gathered from a very short track. Therefore in order to increase the likelihood of successful merges between similar faces, the two similarity measures can be combined (e.g. averaged).
A further optional variation will now be described. When a face stamp is being reconstructed, each block is replaced by a block of the same size, shape and orientation from the reference face stamp. But if the size and orientation of a subject are different in the two face stamps, these face stamps will not be well reconstructed from each other as blocks in the face stamp being reconstructed will not match well with blocks of the same size, shape and orientation. This problem can be overcome by allowing blocks in the reference face stamp to take any size, shape and orientation. The best block is thus chosen from the reference face stamp by using a high order geometric transformation estimation (e.g. rotation, zoom, amongst others). Alternatively the whole reference face stamp can be rotated and resized prior to reconstructing the face stamp by the basic method.
In order to make the similarity measure slightly more robust to lighting variations, each face stamp is first normalised to have a mean luminance of zero and a variance of one.
Use of Face Similarity Component within the Object Tracking System
It has been seen that object tracking allows a person's identity to be maintained throughout a sequence of video frames as long as he/she does not disappear from the scene. The aim of the face similarity component is to be able to link tracks such that the person's identity is maintained even if he/she temporarily disappears from the scene or turns away from the camera, or if the scenes are captured from different respective cameras.
During the operation of the face detection and object tracking system, a new face stamp set is initiated each time a new track is started. The new face stamp set is initially given a unique (i.e. new compared to previously tracked sets) ID. As each stamp of the new face stamp set is acquired, its similarity measure (Si) with all the previously gathered face stamp sets is calculated. This similarity measure is used to update the combined similarity measure (Si−1) of the existing elements of the new face stamp set with all the previously gathered face stamp sets in an iterative manner:
j S i=0.9*j S i−1+0.1*j S i
where the superscript j denotes comparison with the previously gathered face stamp set j.
If the similarity of the new face stamp set to a previously encountered face stamp set is above a certain threshold (T) and the number of elements in the new face stamp set is at least n (see above), then the new face stamp set is given the same ID as the previous face stamp set. The two face stamp sets are then merged to produce just one face stamp set containing as much of the variation contained in the two sets as possible by using the same similarity-comparison method as described in the above section.
The new face stamp set is discarded if its track terminates before n face stamps are gathered.
If the similarity measure of the new face stamp set is above the threshold T for more than one stored face stamp set, this means that the current person appears to be a good match to two previous people. In this case an even more severe similarity threshold (i.e. an even lower difference threshold) is required to match the current person to either of the two previous persons.
In addition to the similarity criterion, another criterion can help in deciding whether two face stamp sets should be merged or not. This criterion comes from the knowledge that two face stamps sets belonging to the same individual cannot overlap in time. Thus two face stamp sets which have appeared in the picture (i.e. from a single camera) at the same time for more than a small number of frames can never be matched to each other. But if they coincide in time but originate from different cameras, the question of matching is slightly more complicated and requires a user to set up the system with camera positions in mind. If the two cameras are directed at the same or similar positions (i.e. so that a person could conceivably be seen by both cameras simultaneously) then matching is allowed. If the two cameras are such that a person could not physically be seen by both simultaneously, then matching of concurrent detections is not allowed. This is all achieved by keeping a record of all the face stamp sets which have ever co-existed in the picture or pictures, using a co-existence matrix. The matrix stores the number of frames for which every combination of two face stamp sets have ever co-existed. If this number is greater than a small number of frames, e.g. 10 frames (to allow for the period during which a track can sometimes float off a face for a few frames prior to being deleted), then it is not permitted to merge the two face stamp sets into the same ID. An example of a coexistence matrix for five people (tracks) numbered ID 1 to ID 5 will now be given. The example concerns only one camera, for simplicity of the explanation, but (as described above) the technique is extendable to multiple cameras.
The matrix shows that:
If two IDs are merged due to high face similarity, the co-existence matrix is updated by combining the co-existence information for the two merged IDs. This is done by simply summing the quantities in the rows corresponding to the two IDs, followed by summing the quantities in the columns corresponding to the two IDs.
e.g. if ID 5 were merged with ID 1, the co-existence matrix above would become:
If ID 1 were subsequently merged with ID 2, the co-existence matrix above would become:
It should be noted that:
In the similarity detection process for generating and for merging face stamp sets, a face stamp typically needs to be reconstructed several times from other face stamps. This means that each block needs to be matched several times using a motion estimation method. For some motion estimation methods, the first step is to compute some information about the block that needs to be matched, irrespective of the reference face stamp used. As the motion estimation needs to be carried out several times, this information can be stored alongside the face stamp, so that it doesn't need to be calculated each time a block has to be matched, thus saving processing time.
The IDs mentioned above can be used (after the matching process) as the basis of the “global face IDs” mentioned earlier.
Although particular embodiments have been described herein, it will be appreciated that the invention is not limited thereto and that many modifications and additions thereto may be made within the scope of the invention. For example, various combinations of the features of the following dependent claims can be made with the features of the independent claims without departing from the scope of the present invention.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US5331544 *||Apr 23, 1992||Jul 19, 1994||A. C. Nielsen Company||Market research method and system for collecting retail store and shopper market research data|
|US5966696 *||Apr 14, 1998||Oct 12, 1999||Infovation||System for tracking consumer exposure and for exposing consumers to different advertisements|
|US6111517 *||Dec 30, 1996||Aug 29, 2000||Visionics Corporation||Continuous video monitoring using face recognition for access control|
|US6185314 *||Feb 6, 1998||Feb 6, 2001||Ncr Corporation||System and method for matching image information to object model information|
|US6246995 *||May 29, 1998||Jun 12, 2001||Ncr Corporation||Product activity data collection system|
|US6263088 *||Feb 6, 1998||Jul 17, 2001||Ncr Corporation||System and method for tracking movement of objects in a scene|
|US6295367 *||Feb 6, 1998||Sep 25, 2001||Emtera Corporation||System and method for tracking movement of objects in a scene using correspondence graphs|
|US6404900 *||Jan 14, 1999||Jun 11, 2002||Sharp Laboratories Of America, Inc.||Method for robust human face tracking in presence of multiple persons|
|US6502076 *||Jun 1, 1999||Dec 31, 2002||Ncr Corporation||System and methods for determining and displaying product promotions|
|US6647269 *||Jul 5, 2001||Nov 11, 2003||Telcontar||Method and system for analyzing advertisements delivered to a mobile unit|
|US6793128 *||Jun 18, 2001||Sep 21, 2004||Hewlett-Packard Development Company, L.P.||Face photo storage system|
|US6795808 *||Oct 30, 2000||Sep 21, 2004||Koninklijke Philips Electronics N.V.||User interface/entertainment device that simulates personal interaction and charges external database with relevant data|
|US7283650 *||Nov 26, 2003||Oct 16, 2007||Video Mining Corporation||Method and system for printing of automatically captured facial images augmented with promotional content|
|US20020072952||Dec 7, 2000||Jun 13, 2002||International Business Machines Corporation||Visual and audible consumer reaction collection|
|US20020075332 *||Sep 22, 1999||Jun 20, 2002||Bradley Earl Geilfuss||Systems and methods for interactive product placement|
|US20020107729 *||Feb 6, 2001||Aug 8, 2002||Catalina Marketing International, Inc.||Method and system for timing promotions based on a prior receipt of promotions|
|US20020178447 *||Apr 3, 2002||Nov 28, 2002||Plotnick Michael A.||Behavioral targeted advertising|
|US20020184098 *||May 8, 2002||Dec 5, 2002||Giraud Stephen G.||Interactive promotional information communicating system|
|US20020190119 *||Jun 18, 2001||Dec 19, 2002||Huffman John W.||Face photo storage system|
|US20030001846 *||Jan 3, 2001||Jan 2, 2003||Davis Marc E.||Automatic personalized media creation system|
|US20030018522 *||Jul 20, 2001||Jan 23, 2003||Psc Scanning, Inc.||Biometric system and method for identifying a customer upon entering a retail establishment|
|US20030044046 *||Aug 28, 2002||Mar 6, 2003||Yoshifumi Nakamura||Method and system for delivering monitored image signal of sbject to be monitored|
|US20030088832 *||Nov 2, 2001||May 8, 2003||Eastman Kodak Company||Method and apparatus for automatic selection and presentation of information|
|US20030208754 *||May 1, 2002||Nov 6, 2003||G. Sridhar||System and method for selective transmission of multimedia based on subscriber behavioral model|
|US20040095604 *||Mar 31, 2003||May 20, 2004||Mark Meyerhofer||Method and apparatus for gaming promotional printer|
|US20040158865 *||Nov 21, 2003||Aug 12, 2004||Kubler Kenneth M.||System and method for managing in-theater display advertisements|
|US20060039733 *||Jun 3, 2005||Feb 23, 2006||Futurelogic, Inc.||Method and apparatus for voucher and coupon printing in a gaming or vending environment|
|WO2001045004A1||Nov 30, 2000||Jun 21, 2001||Promo Vu||Interactive promotional information communicating system|
|WO2004051551A1||Nov 28, 2003||Jun 17, 2004||Sony United Kingdom Limited||Face detection and tracking|
|1||*||Minority Report Advertising (Mini Cooper style), available at http://curtismorley.com/2007/02/06/minority-report-and-mini-cooper/ (describing the technology used in the movie "Minority Report").|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US7987111 *||Oct 26, 2007||Jul 26, 2011||Videomining Corporation||Method and system for characterizing physical retail spaces by determining the demographic composition of people in the physical retail spaces utilizing video image analysis|
|US8218080 *||Jun 6, 2006||Jul 10, 2012||Samsung Electronics Co., Ltd.||Personal settings, parental control, and energy saving control of television with digital video camera|
|US8244005 *||Oct 14, 2010||Aug 14, 2012||Kabushiki Kaisha Toshiba||Electronic apparatus and image display method|
|US8351662||Sep 16, 2010||Jan 8, 2013||Seiko Epson Corporation||System and method for face verification using video sequence|
|US8463052 *||Sep 14, 2010||Jun 11, 2013||Kabushiki Kaisha Toshiba||Electronic apparatus and image search method|
|US8498455||Jun 3, 2010||Jul 30, 2013||Microsoft Corporation||Scalable face image retrieval|
|US8510166 *||May 11, 2011||Aug 13, 2013||Google Inc.||Gaze tracking system|
|US8660307||Mar 29, 2012||Feb 25, 2014||The Nielsen Company (Us), Llc||Methods and apparatus to count people in images|
|US8682388||Dec 31, 2010||Mar 25, 2014||Motorola Mobility Llc||Mobile device and method for proximity detection verification|
|US8712110||Dec 23, 2009||Apr 29, 2014||The Invention Science Fund I, LC||Identifying a characteristic of an individual utilizing facial recognition and providing a display for the individual|
|US8743259 *||Dec 30, 2008||Jun 3, 2014||Nikon Corporation||Information displaying apparatus|
|US8761442||Mar 29, 2012||Jun 24, 2014||The Nielsen Company (Us), Llc||Methods and apparatus to count people in images|
|US8848057||Jun 6, 2006||Sep 30, 2014||Samsung Electronics Co., Ltd.||Home security applications for television with digital video cameras|
|US8860787||May 11, 2011||Oct 14, 2014||Google Inc.||Method and apparatus for telepresence sharing|
|US8862764||Mar 16, 2012||Oct 14, 2014||Google Inc.||Method and Apparatus for providing Media Information to Mobile Devices|
|US8914897||May 23, 2007||Dec 16, 2014||International Business Machines Corporation||Controlling access to digital images based on device proximity|
|US8942994 *||Feb 9, 2012||Jan 27, 2015||Aic Innovations Group, Inc.||Method and apparatus for encouraging consumption of a product|
|US9020833||Aug 23, 2011||Apr 28, 2015||Aic Innovations Group, Inc.||Method and apparatus for social network updates by activity recognition|
|US9092675||Mar 29, 2012||Jul 28, 2015||The Nielsen Company (Us), Llc||Methods and apparatus to count people in images|
|US9129307 *||May 23, 2007||Sep 8, 2015||International Business Machines Corporation||Fee-based distribution of media based on device proximity|
|US9146624||Feb 8, 2012||Sep 29, 2015||Google Technology Holdings LLC||Method for managing screen orientation of a portable electronic device|
|US9158970||Nov 11, 2013||Oct 13, 2015||Canon Kabushiki Kaisha||Devices, systems, and methods for visual-attribute refinement|
|US9275285||Mar 29, 2012||Mar 1, 2016||The Nielsen Company (Us), Llc||Methods and apparatus to count people in images|
|US9292736||Dec 20, 2013||Mar 22, 2016||The Nielsen Company (Us), Llc||Methods and apparatus to count people in images|
|US9430719 *||Sep 16, 2015||Aug 30, 2016||Google Inc.||System and method for providing objectified image renderings using recognition information from images|
|US9465999||May 19, 2014||Oct 11, 2016||The Nielsen Company (Us), Llc||Methods and apparatus to count people in images|
|US9479274||Sep 27, 2007||Oct 25, 2016||Invention Science Fund I, Llc||System individualizing a content presentation|
|US9530144 *||May 23, 2011||Dec 27, 2016||Rakuten, Inc.||Content output device, content output method, content output program, and recording medium having content output program recorded thereon|
|US9582724 *||Jan 27, 2016||Feb 28, 2017||Microsoft Technology Licensing, Llc||False face representation identification|
|US9594961||Jun 5, 2015||Mar 14, 2017||The Nielsen Company (Us), Llc||Methods and apparatus to count people in images|
|US9628552||Oct 2, 2014||Apr 18, 2017||Google Inc.||Method and apparatus for digital media control rooms|
|US9633186 *||Apr 23, 2012||Apr 25, 2017||Apple Inc.||Systems and methods for controlling output of content based on human recognition data detection|
|US9647780 *||Aug 24, 2007||May 9, 2017||Invention Science Fund I, Llc||Individualizing a content presentation|
|US20070126873 *||Jun 6, 2006||Jun 7, 2007||Samsung Electronics Co., Ltd.||Home security applications for television with digital video cameras|
|US20070126884 *||Jun 6, 2006||Jun 7, 2007||Samsung Electronics, Co., Ltd.||Personal settings, parental control, and energy saving control of television with digital video camera|
|US20070219866 *||Mar 13, 2007||Sep 20, 2007||Robert Wolf||Passive Shopper Identification Systems Utilized to Optimize Advertising|
|US20080294548 *||May 23, 2007||Nov 27, 2008||David Keith Fowler||Fee-Based Distribution of Media Based on Device Proximity|
|US20080294774 *||May 23, 2007||Nov 27, 2008||David Keith Fowler||Controlling Access to Digital Images Based on Device Proximity|
|US20090051542 *||Aug 24, 2007||Feb 26, 2009||Searete Llc, A Limited Liability Corporation Of The State Of Delaware||Individualizing a content presentation|
|US20090180004 *||Dec 30, 2008||Jul 16, 2009||Nikon Corporation||Information displaying apparatus|
|US20110064311 *||Sep 14, 2010||Mar 17, 2011||Kabushiki Kaisha Toshiba||Electronic apparatus and image search method|
|US20110074540 *||Oct 29, 2009||Mar 31, 2011||Hon Hai Precision Industry Co., Ltd.||Control system and method for interface of electronic device|
|US20110110564 *||Oct 14, 2010||May 12, 2011||Kabushiki Kaisha Toshiba||Electronic apparatus and image display|
|US20110119111 *||Nov 16, 2009||May 19, 2011||Global Rainmakers, Inc.||Measuring Effectiveness of Advertisements and Linking Certain Consumer Activities Including Purchases to Other Activities of the Consumer|
|US20110148753 *||Dec 23, 2009||Jun 23, 2011||Searete Llc, A Limited Liability Corporation Of The State Of Delaware||Identifying a characteristic of an individual utilizing facial recognition and providing a display for the individual|
|US20110150276 *||Dec 23, 2009||Jun 23, 2011||Searete Llc, A Limited Liability Corporation Of The State Of Delaware||Identifying a characteristic of an individual utilizing facial recognition and providing a display for the individual|
|US20110150294 *||Dec 23, 2009||Jun 23, 2011||Searete Llc, A Limited Liability Corporation Of The State Of Delaware|
|US20110150295 *||Dec 23, 2009||Jun 23, 2011||Searete Llc, A Limited Liability Corporation Of The State Of Delaware|
|US20110150296 *||Dec 23, 2009||Jun 23, 2011||Searete Llc, A Limited Liability Corporation Of The State Of Delaware|
|US20110150297 *||Dec 23, 2009||Jun 23, 2011||Searete Llc, A Limited Liability Corporation Of The State Of Delaware|
|US20110150298 *||Dec 23, 2009||Jun 23, 2011||Searete Llc, A Limited Liability Corporation Of The State Of Delaware|
|US20110150299 *||Dec 23, 2009||Jun 23, 2011||Searete Llc, A Limited Liability Corporation Of The State Of Delaware|
|US20110206245 *||Jan 25, 2011||Aug 25, 2011||Searete Llc, A Limited Liability Corporation Of The State Of Delaware|
|US20110211738 *||Jan 25, 2011||Sep 1, 2011||Searete Llc, A Limited Liability Corporation Of The State Of Delaware|
|US20110211739 *||Jan 25, 2011||Sep 1, 2011||Searete Llc, A Limited Liability Corporation Of The State Of Delaware|
|US20110238361 *||Sep 4, 2008||Sep 29, 2011||Kazuya Ueki||Tallying system, tallying apparatus and tallying method|
|US20120239458 *||Nov 16, 2009||Sep 20, 2012||Global Rainmakers, Inc.||Measuring Effectiveness of Advertisements and Linking Certain Consumer Activities Including Purchases to Other Activities of the Consumer|
|US20120290401 *||May 11, 2011||Nov 15, 2012||Google Inc.||Gaze tracking system|
|US20130067513 *||May 23, 2011||Mar 14, 2013||Rakuten, Inc.||Content output device, content output method, content output program, and recording medium having content output program recorded thereon|
|US20130211889 *||Feb 9, 2012||Aug 15, 2013||Ai Cure Technologies Inc.||Method and apparatus for encouraging consumption of a product|
|US20130279744 *||Apr 23, 2012||Oct 24, 2013||Apple Inc.||Systems and methods for controlling output of content based on human recognition data detection|
|US20140244385 *||Feb 26, 2014||Aug 28, 2014||Kt Corporation||Advertisement service using mobile vehicle|
|US20160125270 *||Sep 16, 2015||May 5, 2016||Google Inc.||System And Method For Providing Objectified Image Renderings Using Recognition Information From Images|
|US20160140406 *||Jan 27, 2016||May 19, 2016||Microsoft Technology Licensing, Llc||False face representation identification|
|US20160292713 *||Mar 31, 2015||Oct 6, 2016||Yahoo! Inc.||Measuring user engagement with smart billboards|
|U.S. Classification||382/118, 382/103, 705/7.29, 705/14.41|
|International Classification||G06Q30/00, G06K9/00, G09F27/00, G06T1/00, G06F17/30|
|Cooperative Classification||G06K9/00295, G06K9/00248, G06K2009/00322, G09F27/00, G06Q30/0201, G06K9/00221, G06Q30/02, G06Q30/0242, G06K9/00771|
|European Classification||G06Q30/02, G06Q30/0242, G06Q30/0201, G06K9/00F, G06K9/00F3U, G09F27/00, G06K9/00V4, G06K9/00F1L|
|May 2, 2005||AS||Assignment|
Owner name: SONY UNITED KINGDOM LIMITED, UNITED KINGDOM
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:KILNER, ANDREW ROGER;JENNINGS, VICTORIA SOPHIA;PASZKOWICZ, SEBASTIAN ALEKSANDER;AND OTHERS;REEL/FRAME:016508/0795;SIGNING DATES FROM 20050102 TO 20050127
|Mar 14, 2013||FPAY||Fee payment|
Year of fee payment: 4
|Jun 26, 2017||FPAY||Fee payment|
Year of fee payment: 8